BPN14770

BPN14770 was designed at Tetra Discovery Partners in Grand Rapids, Michigan, later Tetra Therapeutics and now a wholly owned subsidiary of Shionogi, and has since been given the name zatolmilast [1][7]. It addresses a long-standing problem. Inhibitors of PDE4, the main enzyme that breaks down cAMP in neurons, improve memory in animals, but nausea and vomiting stop the dose being raised far enough to test the idea properly in people [2][3]. BPN14770 inhibits only the PDE4D subtype, only partly, and mainly when the enzyme has been switched on by cAMP signalling [2].

The mouse pharmacology needed a special mouse. The pocket the drug uses differs by one amino acid between primates and every other mammal, so the group changed a single codon in mice to give them the human version [3]. In those animals one oral dose of 0.01–0.03 mg/kg improved Y-maze alternation and 24-hour object recognition, about 100 times less than ordinary littermates needed, and 0.03 mg/kg reversed scopolamine-induced amnesia [3]. The same laboratory then showed that 14 days at 0.01–0.03 mg/kg prevented the memory loss, dendrite loss and synaptic-protein loss caused by injecting amyloid-β into the hippocampus [7], and reported a further scopolamine study [8]. It had no effect in tests of antidepressant-like or anxiolytic-like behaviour [3]. All three papers come from one laboratory at the University at Buffalo, with a Tetra employee as co-author.

Fragile X syndrome became the lead indication because cAMP production is low in patients' cells and in fly and mouse models of the disorder [9][10]. In adult fmr1-knockout mice, 0.3 mg/kg a day for two weeks brought open-field hyperactivity, social sniffing, nesting and marble burying back to the level of normal mice, left normal mice unchanged, and shortened the over-long dendritic spines of cortical layer 3 neurons by about 10%; the behavioural benefit was still there two weeks after the last dose [9]. A National Institute of Mental Health group, again with Tetra's co-authorship, fed it to knockout mice from weaning. Sleep and social behaviour increased in knockout and normal mice alike and open-field hyperactivity normalised at the lower doses, but activity in the elevated plus maze did not change and effects on brain protein synthesis were unclear [11].

Independent groups have used it as a tool. In ordinary mice, 0.1 mg/kg given twice during sleep deprivation prevented the resulting deficits in spatial pattern separation and object-location memory, with no effect in rested mice [12][13]. A Korean laboratory found none of the anti-inflammatory action of the non-selective PDE4 inhibitor roflumilast [14].

The human record is one small positive trial and three larger negative ones.

Phase 2, fragile X syndrome. Thirty men at a single centre took 25 mg twice daily and placebo for 12 weeks each, with no washout in between [4]. The benefit persisted after the switch to placebo, and low drug levels were still measurable 12 weeks after the last dose, so under a rule written into the analysis plan the efficacy results rest on the first period alone, 15 men against 15 [4]. In that comparison the NIH Toolbox crystallised cognition composite favoured the drug by 5.3 points (P = 0.0018), as did its two components, picture vocabulary and oral reading, and caregivers' ratings of language and of daily functioning each favoured it by 14 points on a 100-point scale [4]. Part of the cognitive gap came from a fall in the placebo group, −3.1 against +2.2, which the authors acknowledge [4]. Most other measures did not reach significance: memory and processing-speed tests, six of seven attention-test scores, caregiver ratings of anxiety and irritability, a behaviour checklist, a mood scale, adaptive behaviour, which was numerically slightly worse on the drug, and clinicians' global ratings. No correction was made for the number of comparisons [4]. The figures posted on the registry, which count all 30 men under each treatment, show much smaller gaps: a composite change of +0.1 against −0.9, and caregiver ratings less than 3 points apart [15].

Phase 2, early Alzheimer's disease. In 255 people on a cholinesterase inhibitor, 10 or 25 mg twice daily for 13 weeks gave the same result as placebo. Delayed memory scores rose by 10.1, 9.9 and 9.7 points in the three groups, and the secondary cognitive, functional and global scores were just as close. The registry posts the group means without a statistical comparison [5].

Phase 3, fragile X syndrome. A 13-week trial in 171 men missed its primary endpoint, the same NIH Toolbox composite. The companion trial in 163 boys aged 9–17 missed its primary endpoints, caregiver ratings of language and daily function, and its key secondary endpoint [6][16][17]. Shionogi reports significant differences on the caregiver ratings in the adult trial, a secondary measure there, and calls them suggestive of a signal but not conclusive. It has released no numbers [6].

PDE4D has a lid, and the drug holds it shut. The PDE4 enzymes carry two regulatory regions upstream of the catalytic site. One, UCR1, lets the long forms of the enzyme pair into dimers and is where PKA phosphorylates and activates them; the other, UCR2, is a helix that swings over the active site and controls access by cAMP [2][3]. Classical inhibitors such as apremilast sit in the active site, which is identical in all four PDE4 subtypes, so they cannot tell the subtypes apart [2]. BPN14770 also sits in the active site but completes a surface that lets UCR2 close down on top of it [3][18]. UCR2 carries a phenylalanine in PDE4D where PDE4A, B and C have a tyrosine, and the drug is shaped to accommodate the first and clash with the second [2][3].

Three kinds of selectivity follow. It inhibits activated long-form PDE4D at 7–8 nM, about 260 times more potently than PDE4B. It is 16 times weaker against a short, monomeric form of PDE4D that lacks the dimerising region, and about 130 times weaker against the long form before PKA has activated it [3]. So it acts chiefly where cAMP signalling is already under way, which the developers argue preserves the timing and location of the signal instead of flooding the cell [2].

Inhibition is deliberately incomplete. In the dimer, closing one active site slows the other without shutting it, so inhibition levels off at about 90% where apremilast reaches 100% [2][18].

The species difference is real and large. That phenylalanine exists only in primates. The mouse enzyme is inhibited at 133 nM; swapping in the human residue brings that to 2.9 nM and creates a high-affinity binding site in mouse brain that was not there before [3]. This is why results in ordinary rodents understate the drug, and why its low-dose data all come from one engineered mouse line.

Downstream it follows the textbook memory pathway. In humanised mice it raised brain cAMP, let a weak stimulus produce long-lasting potentiation in hippocampal slices at 100 nM, and after two weeks raised CREB phosphorylation, BDNF, phosphorylated synapsin and PSD95 in the hippocampus [3]. A PKA inhibitor infused into the brain abolished the memory benefit [3][7]. One paper adds SIRT1 and Akt as necessary intermediates [8].

Why vomiting should be less of a problem. Mice cannot vomit, so the group used a stand-in, the shortening of ketamine–xylazine anaesthesia. In humanised mice that began at 3 mg/kg, against memory effects at 0.01–0.03 mg/kg, a 40- to 100-fold gap in plasma exposure [3]. Ferrets did not vomit at up to 30 mg/kg, nor monkeys at up to 50 mg/kg [2]. The authors' explanation is that the brainstem's vomiting centre mostly expresses short PDE4D forms the drug binds poorly [3].

Human genetics supplied the rationale. Rare missense mutations in PDE4D cause acrodysostosis with intellectual disability, and common variants over the exons encoding the long forms are associated with cognitive performance in genome-wide studies [3][19][20][21].

In people, target engagement is partly shown. A single 50 mg dose reduced binding of a PDE4D PET tracer in the brain, by roughly 40%, though tracer metabolites made the figure unreliable [22]. In the fragile X trial its effect on the auditory N1 response, a marker of cortical over-excitability in the syndrome, was a non-significant trend [4]. A later analysis found that men with higher blood levels at the end of the first period had smaller N1 responses, and a machine-learning reanalysis picked out a rise in peak alpha frequency on the drug [1][23]. Both come from 12 to 23 participants and both sets of authors ask for replication.

BPN14770 is a 406 Da pyridine bearing a phenylacetic acid arm. It was reached from an earlier amide lead whose half-life in mice was under half an hour: the amide was replaced with a carboxylic acid and a trifluoromethyl group was added to block oxidation of the ring's side chain [2]. Carboxylic acids had been inactive in the project's earlier chemical series, so its potency came as a surprise to its makers [2]. It dissolves to 95 µM at pH 7.4, crosses cell monolayers without being pumped back by P-glycoprotein, and is 99.5–99.8% bound to plasma proteins [2][14].

The phase 2 trial used 25 mg capsules, and the adolescent phase 3 trial gave 15 or 25 mg twice daily [15][17]. Animal studies have used a suspension in 0.5% methylcellulose, a Solutol solution, medicated chow, and a solvent mix of N-methylpyrrolidone and polyethylene glycol [8][9][11][14].

How well it is absorbed is not settled. The developer reports oral bioavailability of 70–82% in mice, 100% from the methylcellulose suspension, and 80–100% in rats and dogs [2][3][9]. The one independent measurement, in mice with the solvent mix, was 3.7%, and that group concluded brain penetration was poor [14]. The two have not been reconciled.

Tolerability in trials has been good, with a mild gastrointestinal signal. In the Alzheimer's trial, the largest with posted data, serious adverse events occurred in two people in each of the three arms and nobody died. Nausea affected 4 of 80 on 10 mg and 5 of 85 on 25 mg against none of 86 on placebo, and diarrhoea 5, 6 and 3 [5]. The serious events on the drug were cholecystitis, rhabdomyolysis, presyncope and a hip fracture, one each [5]. In the fragile X phase 2 trial all 30 men completed both periods, most of them while staying on their usual medicines, which included SSRIs, methylphenidate and antipsychotics. Eleven had an adverse event while on the drug and 8 while on placebo, and vomiting occurred in 3 and 2. The investigator judged no event to be related to treatment, including the one serious event, a septic elbow bursitis that came with a transient rise in liver enzymes. ECGs did not change [4][15]. Shionogi describes the phase 3 trials as generally well tolerated with no new safety concerns, without figures [6].

The dose is capped by nausea. The phase 1 dose-escalation studies in healthy volunteers found the threshold for nausea and vomiting at a peak plasma level of about 1,500 ng/mL, and the 25 mg twice-daily regimen was chosen to stay under it [4]. Headache was the most frequent complaint in elderly volunteers on the highest phase 1 dose [26].

Animal toxicology is reported in summary. Rats showed no central nervous system or lung effects at up to 60 mg/kg and dogs no cardiovascular effect at up to 100 mg/kg. At 100 mg/kg rats lost appetite and weight, and dogs vomited, trembled and lost weight [2]. Studies of 28 and 90 days in rats and dogs found no vascular toxicity, the lesion that non-selective PDE4 inhibitors cause in animals, and no other microscopic change [2]. It did not block the hERG channel below 10 µM and was not mutagenic in bacteria [2].

What is missing. The phase 3 safety data and the long-term extension have not been published. The fragile X trials enrolled only males [15][16][17].